AG2PI Workshop #14 - Tuesday July 26, 2022
Intermediate Omics Data-Enabled Genomic Prediction and Mediation Analysis
Tuesday July 26, 2022
12:00 - 2:00 PM
(US Central Time)
Purpose
Learn the basic principles of applying and integrating Omics data to predict genetic loci of agronomic traits.
Registration
Register for the virtual event by clicking the link below. Upon registration, you will receive a confirmation email with information about joining the meeting
Workshop RegistrationWorkshop Resources
Click the buttons below to access resources and/or watch recording of this workshop
Watch Workshop Recording Github Repository Google Colab NotebookChat Questions
Questions directed to the speakers placed in the chat can be viewed by clicking the button below
See Chat QuestionsThrough this two-hour workshop, participants will learn the basic principle of establishing the successive relationship from SNP to intermediate Omics data and then to conventional phenotypes. Participants will then get hands-on experience integrating population-wide Omics data to identify mediator genes and predict breeding values.
About Presenters
Dr. Jinliang Yang is an assistant professor at the University of Nebraska-Lincoln. His group focuses on quantitative genetics and statistical genomics of maize and its wild relatives, from historical domestication to future crop improvement.
Dr. Hao Cheng is an assistant professor of quantitative genetics in the Deparment of Animal Science at the University of California, Davis. His research interests are broadly involved in the development of statistical and computational methods for the (genetic) improvement of populations through more accurate, efficient, and biologically meaningful analysis. His lab has focused on the use of genomics, phenomics, pedigree, and other sources of big data in various species to better predict desired traits.
Chat Questions
You described what one may call 1-step mediator approach, but have you tested associations with perhaps multiple mediated steps (ie. like 2-step where Mediator1 may associate with a Mediator2 that then associates with Phenotype)? It would become a more complex model. Would that be possible or would it lead to too more noise?
It is a great idea. Yes, what we talked about here is indeed a 1-step mediation approach. If you consider one gene's expression (i.e., measured using RNA-seq read count) as the outcome (or phenotype), similar to the eQTL study, you can simply construct a model SNP mediators other than gene A gene A
. It is still a 1-step mediation model
, but incorporates two Omics layers. I haven't tested this idea yet. For a model with a longer chain, I assume it would be too noisy.
I think this integrated approach (GMA) can be used also for recognition of core and peripheral genes that described by the Boyle et al. Omnigenic Model, alongside to response to some part of hidden heritability of traits, is it not so?
I agree with you, GMA can estimate variance explained by direct effect pathway (conventional heritability) and the indirect effect pathway (additional variance explained by the mediation processes)
How does one include other fixed effects like farm, pen, feed/diet etc? Or do the phenotypes need to be pre-corrected for these?
Users can specify other fixed effects in the X0 matrix.
The point of doing GMA analysis is to identify causal/likely-causal genes of some trait? Do you also do genome predictions, and will it improve prediction accuracy? Or this will be more of Dr. Cheng's research topics.
Currently, the GMA method focuses only on inference, not on prediction.
did you say that the plots won't work if I have more chromosomes?
The plotting method has been updated to work for species with any number of chromosomes.
